clang claims weak aliases are not supported on Darwin but only when using __attribute__, works fine with #pragma

[CodingMarkus]

clang claims weak aliases are not supported on Darwin. The following code doesn't compile:

#include <stdio.h>

int func1( ) { return 42; }
__attribute__((weak, alias("func1")))  int func2( );

int main( )
{
	printf("%d\n", func2());
	return 0;
}

Output:

test1.c:4:22: error: aliases are not supported on darwin
__attribute__((weak, alias("func1")))  int func2( );
                     ^
1 error generated.

But that's not true. Weak aliases are supported on Darwin. The following code will compile and will behave correctly:

#include <stdio.h>

int func1( ) { return 42; }

#pragma weak func2 = func1
int func2( );

int main( )
{
	printf("%d\n", func2());
	return 0;
}

[CodingMarkus]

This could be seen as a dupe of #11488, albeit that report was about the error message which used to be only weak aliases are supported on darwin at that time and made even less sense to me, since it came when the code was trying to make exactly that, a weak alias.

[shafik]

CC @rjmccall

[llvmbot]

@llvm/issue-subscribers-clang-frontend

Author: None (CodingMarkus)

clang claims weak aliases are not supported on Darwin. The following code doesn't compile:

#include <stdio.h>

int func1( ) { return 42; }
__attribute__((weak, alias("func1")))  int func2( );

int main( )
{
	printf("%d\n", func2());
	return 0;
}

Output:

test1.c:4:22: error: aliases are not supported on darwin
__attribute__((weak, alias("func1")))  int func2( );
                     ^
1 error generated.

But that's not true. Weak aliases are supported on Darwin. The following code will compile and will behave correctly:

#include <stdio.h>

int func1( ) { return 42; }

#pragma weak func2 = func1
int func2( );

int main( )
{
	printf("%d\n", func2());
	return 0;
}